On the eve of COMPUTEX 2025 in Taipei, Intel unveiled a series of graphics processing units (GPUs) and AI accelerators tailored for professional creators, AI developers, and enterprise users, showcasing its latest布局 in artificial intelligence, open computing platforms, and professional graphics. The announcements center on three core pillars: the new generation of Intel Arc Pro B-series GPUs, expanded deployment options for the Gaudi 3 AI accelerator, and a new open-source framework called “Intel AI Assistant Builder.”
Welcoming the New AI Era: Arc Pro B-Series GPUs Designed for Workstations and Inference Tasks
Amid the ongoing wave of AI and high-performance computing (HPC) applications, Intel has unveiled the new Intel® Arc™ Pro B60 and B50 graphics chips based on the Xe2 architecture. These two professional-grade GPUs mark a further expansion of Intel’s Arc Pro product line, delivering high performance for creators, developers, and engineers. The Arc Pro B60 comes with 24GB of memory, while the Arc Pro B50 offers 16GB, both integrating Intel Xe Matrix Extensions (XMX) AI engines and ray tracing units. This not only enhances graphics rendering quality but also provides hardware acceleration for AI workloads, further strengthening their value in professional fields such as architecture, engineering, and construction (AEC).
According to Intel’s technical documentation, the B-series GPUs support multi-GPU scalability and can be flexibly configured based on task scale, providing flexibility for small and medium-sized businesses or large workstation deployments. The products have also received certifications from multiple independent software vendors (ISVs), and optimized software delivers stability and performance.Intel Arc Pro BThe series is atWindowsDrivers on the system that are compatible with consumer and professional editions; at
Intel also released configurable workstation-gradeIntel® Xeon®Architecture Platform (codenameProject Battlematrix), is specifically designed to reduceAIDesigned for developers’ pain points
Intel Vice President and General Manager of the Client Graphics Group, Vivian Lien, stated: “Intel Arc Pro BThe series demonstrates Intel’s commitment to display chip technology and ecosystem partnerships. ThroughX
The Arc Pro B60 is expected to be available for sampling from June through partners including ASRock, Lanner, GUNNIR, and Senao; the Arc Pro B50 is expected to go on official sale through authorized channels starting in July.
Gaudi 3 AI Accelerator: A Comprehensive AI Scaling Solution from PCIe to Rack-Scale
As AI model scales continue to expand, the demand for high-performance and flexible hardware architectures is growing increasingly urgent. Intel’s newly announced Gaudi 3 AI accelerator offers two primary deployment options: a PCIe card version and a rack-level system reference design, addressing AI inference and training needs across enterprises of different sizes. The Gaudi 3 PCIe card can be deployed directly into server racks in existing data centers, making it suitable for small and medium-sized enterprises looking to upgrade their AI capabilities. This accelerator card supports inference workloads ranging from small models like Llama 3.1 8B to advanced large models such as Llama 4 Maverick, providing developers with a modular, cost-effective solution. The product is expected to be officially available in the second half of 2025.
Furthermore, the next-generation Gaudi 3 rack-scale system design supports up to 64 accelerators per rack and a total of 8.2 TB of high-bandwidth memory. Liquid cooling technology maintains efficient heat dissipation and energy-saving performance, making it suitable for cloud service providers (CSPs) and research institutions to deploy large models. Its open modular architecture helps avoid vendor lock-in, echoing Intel’s long-term commitment to open, flexible, and scalable AI architectures, and is also compatible with the Open Compute Project (OCP).
Intel AI Assistant Builder: Open-Source Tool Enables Local Deployment of AI Agents
After its debut at CES 2025, Intel officially open-sourced its lightweight AI agent framework, “Intel AI Assistant Builder,” and published it on GitHub for developers to use free of charge. The framework supports local operation on AI PCs powered by Intel processors, helping enterprises and individuals rapidly develop and deploy custom AI assistant applications. This tool is especially well-suited for laptop manufacturers, system integrators, and software companies looking to create tailored AI experiences with their own products. At COMPUTEX, Acer and ASUS have already showcased innovative use cases built with AI Assistant Builder.
Intel emphasizes that this tool will drive the expansion of applications in the AI PC market, helping developers integrate AI features such as voice assistants, document summarization, and image recognition with a lower barrier to entry, while also preserving the privacy and low-latency advantages of on-device operation.
Celebrating Intel’s 40 Years in Taiwan: From Chip Manufacturing to Key AI Ecosystem Partner
This launch event coincides with Intel’s 40th anniversary of operations in Taiwan, symbolizing its history and outlook of close collaboration with Taiwan’s industrial ecosystem. From the early x86 processor supply chain to today’s comprehensive deployment covering AI PCs, data centers, and cloud acceleration platforms, Intel’s technological collaboration with Taiwan remains ever-fresh and enduring.
“For the past forty years, the vitality and innovation of Taiwan’s technology industry have driven Intel forward. We will continue to deepen our collaboration with local partners to jointly develop products and services that meet the demands of future markets.”
Source: KOCPC Chinese